3
$\begingroup$

We believe that if after running blast, the global identity between a resulting sequence and our query is at least 30%, we can say that two sequences are homologs. what is the difference between local identity and global identity and how can we calculate them? the following file is the global alignment result of one of my PSI-BLAST hits against query:

Aligned_sequences: 2
# 1: NP_418577.1
# 2: WP_036312822.1
# Matrix: EBLOSUM62
# Gap_penalty: 10.0
# Extend_penalty: 0.5
#
# Length: 264
# Identity:      85/264 (32.2%)
# Similarity:   125/264 (47.3%)
# Gaps:          29/264 (11.0%)
# Score: 376.5
# 
#
#=======================================

NP_418577.1        1 -----------MAEMKNLKIE-------VVRYNPEVDTAPHSAFYEVPYD     32
                                ..|..:..|:       :.|::||||..|....|:|...
WP_036312822.      1 MTIVDSGAPADTQEANDSGIQSYLVTFIIRRFDPEVDAEPRWVDYDVEMY     50

NP_418577.1       33 ATTSLLDALGYIKDNLAPDLSYRWSCRMAICGSCGMMVNNVPKLACKTFL     82
                     .|..:||||..||.::...||:|.||...||||..|.:|...:|||||.:
WP_036312822.     51 PTDRVLDALHRIKWDVDGTLSFRRSCAHGICGSDAMRINGRNRLACKTLI    100

NP_418577.1       83 R--DYTDGMKVEALANFPIERDLVVDMTHFIESLEAIKPYIIGNSRTADQ    130
                     :  |.:..:.|||:...|:|:||:|||..|.||...::|::...|.....
WP_036312822.    101 KDLDISKPIYVEAIKGLPLEKDLIVDMDPFFESFRDVQPFLQPKSAPEPG    150

NP_418577.1      131 GTNIQTPAQMAKYHQFSGCINCGLCYAACPQFGLNPEFIGPAAITLAHRY    180
                     ....|:....|.|...:.||.|..|.::||.|..:.::.|||||..|||:
WP_036312822.    151 KERFQSIKDRAVYDDTTKCILCAACTSSCPVFWTDGQYFGPAAIVNAHRF    200

NP_418577.1      181 NEDSRDHGKKERMAQLNSQNGVWSCTFVGYCSEVCPKHVDPAAAIQQGKV    230
                     ..||||.....|:..||.:.|||.|.....|:|.||:.::...||.:.|.
WP_036312822.    201 IFDSRDDAADVRLDILNDKEGVWRCRTTFNCTEACPRGIEITKAIAEVKQ    250

NP_418577.1      231 ESSKDFLIATLKPR    244
                     ...:.         
WP_036312822.    251 AVLRG---------    255


#---------------------------------------
#---------------------------------------
the following file is the local alignment of that hit against query:
Aligned_sequences: 2
# 1: NP_418577.1
# 2: WP_036312822.1
# Matrix: EBLOSUM62
# Gap_penalty: 10.0
# Extend_penalty: 0.5
#
# Length: 219
# Identity:      83/219 (37.9%)
# Similarity:   119/219 (54.3%)
# Gaps:           2/219 ( 0.9%)
# Score: 390.5
# 
#
#=======================================

NP_418577.1       13 RYNPEVDTAPHSAFYEVPYDATTSLLDALGYIKDNLAPDLSYRWSCRMAI     62
                     |::||||..|....|:|....|..:||||..||.::...||:|.||...|
WP_036312822.     31 RFDPEVDAEPRWVDYDVEMYPTDRVLDALHRIKWDVDGTLSFRRSCAHGI     80

NP_418577.1       63 CGSCGMMVNNVPKLACKTFLR--DYTDGMKVEALANFPIERDLVVDMTHF    110
                     |||..|.:|...:|||||.::  |.:..:.|||:...|:|:||:|||..|
WP_036312822.     81 CGSDAMRINGRNRLACKTLIKDLDISKPIYVEAIKGLPLEKDLIVDMDPF    130

NP_418577.1      111 IESLEAIKPYIIGNSRTADQGTNIQTPAQMAKYHQFSGCINCGLCYAACP    160
                     .||...::|::...|.........|:....|.|...:.||.|..|.::||
WP_036312822.    131 FESFRDVQPFLQPKSAPEPGKERFQSIKDRAVYDDTTKCILCAACTSSCP    180

NP_418577.1      161 QFGLNPEFIGPAAITLAHRYNEDSRDHGKKERMAQLNSQNGVWSCTFVGY    210
                     .|..:.::.|||||..|||:..||||.....|:..||.:.|||.|.....
WP_036312822.    181 VFWTDGQYFGPAAIVNAHRFIFDSRDDAADVRLDILNDKEGVWRCRTTFN    230

NP_418577.1      211 CSEVCPKHVDPAAAIQQGK    229
                     |:|.||:.::...||.:.|
WP_036312822.    231 CTEACPRGIEITKAIAEVK    249


#---------------------------------------
#---------------------------------------
Improve this question
$\endgroup$
8
  • 4
    $\begingroup$ BLAST (Basic Local Alignment Search Tool) doesn't give you global alignments, so you may need to reconsider what you're doing/asking. Furthermore, "homolog" has a very specific definition. Homologs are genes that are both descended from the same ancestral sequence. It doesn't, as many often mistake, mean 'similarity'. 30% sequence identity, does not imply homology by itself. You need more information. $\endgroup$
    – Joe Healey
    Jun 21, 2017 at 8:49
  • $\begingroup$ Take a look here (ebi.ac.uk/Tools/psa) if you want to find tools for local and global alignment. $\endgroup$
    – Joe Healey
    Jun 21, 2017 at 8:50
  • 2
    $\begingroup$ Perhaps I'm wrong but "They must have globally good identity" still isn't really enough to say that these sequences are homologs. You need ancestry information for the organisms you're interested in. If you really want to know orthology/homology, I'd suggest using a dedicated orthology finding program like OrthAgogue, OrthoMCL, or better yet ROARY if your sequences are bacterial. $\endgroup$
    – Joe Healey
    Jun 21, 2017 at 9:04
  • 1
    $\begingroup$ No they don't. It all depends on how closely related the species are. Please edit your question and add more information. There is no magix sequence simlarity threshold that defines homology. $\endgroup$
    – terdon ♦
    Jun 21, 2017 at 9:04
  • 2
    $\begingroup$ There is no magic formula. For closely related species I would expect a sequence identity of far more than 30% for a true homolog but for distant species, I could also expect far lower. As I told you in my answer to your previous question, it is far more important to look at the conservation of specific residues than to look at the overall sequence similarity value. $\endgroup$
    – terdon ♦
    Jun 21, 2017 at 9:15

2 Answers 2

Reset to default
3
$\begingroup$

If you want a purely alignment based approach to begin to infer homology (note I say infer, as we stated in the comments, this isn't sufficient to say 2 sequences are homologs), I would use a Multiple Sequence Alignment (MSA) instead of a pairwise like you've done so far, and include as many similar proteins as you can from BLAST.

The choice of algorithm for the MSA probably won't matter too much, as you will be looking for conserved regions anyway.

Once you have your MSA, you can easily spot conserved domains. If anything is known about this protein structurally, it's quite trivial to work out whether a conserved region is a functional domain (active site, that kind of thing).

With this information, you would be somewhat justified in calling these proteins structural homologs at least if they share domains, but, to reiterate, without doing phylogenetic analyses, you cannot call them true homologs.

Improve this answer
$\endgroup$
1
$\begingroup$

Defining homology isn't as simple as an identity cutoff, however, you're right that homology seems likely given a certain identity. This is where e-value comes into play. See: https://en.wikipedia.org/wiki/Expected_value

To answer the question, global-global requires that the alignment account for every base in the query and the target sequence. local-global requires a the full length of the query on a part of the target. local-local allows only the highest scoring part of the alignment between the query and target to be reported. i.e. one functional domain against another. these hits are often called HSPs.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.